A Spectral Interpretable Bearing Fault Diagnosis Framework Powered by Large Language Models
Abstract
1. Introduction
2. Related Works
2.1. Bearing Fault Diagnosis Techniques
2.2. Interpretable Artificial Intelligence in Fault Diagnosis
2.3. Large Language Models for Explanation and Reasoning
3. Methodology
Algorithm 1 Overall Workflow of the Spectral Interpretable Diagnosis Framework |
|
3.1. Data Preprocessing
3.2. Fault Classification Network
3.3. Interpretable Diagnosis with LLMs
4. Experimental Study
4.1. Experimental Setup
4.2. Fault Classification Results
4.3. Analysis of Interpretability
4.4. Evaluation of Diagnostic Interpretability and Reasoning
- FCN+Mapping: This baseline method utilizes the fault classification probabilities from the FCN, which are then directly mapped to predefined textual labels as the diagnostic output. This represents a common approach in automated systems where classification results are translated into simple reports without elaborate reasoning.
- Ours (w/o FCN): An ablation of our proposed method where the FCN module is omitted. The fine-tuned LLM directly analyzes the input spectral data to generate a diagnosis. This configuration is designed to evaluate the LLM’s standalone analytical capability on spectral features without the FCN’s initial probabilistic guidance, which, as hypothesized, might lead to lower diagnostic accuracy due to the absence of focused fault categorization.
- Ours (w/o LoRA): Another ablation of our method where the LLM (Qwen3-4B) is used without the LoRA fine-tuning. In this setup, the LLM processes both the raw spectral information and the FCN’s output probabilities to generate a diagnosis. This configuration aims to highlight the impact of domain-specific fine-tuning on the quality and relevance of the LLM’s generated explanations.
- ChatGPT (GPT-4 class API-based LLM) [36]: Representing advanced, general-purpose large language models accessible via API. While capable of high-quality text generation and complex understanding, its application here is without specific fine-tuning for the bearing fault diagnosis task. Potential drawbacks for industrial deployment include inference costs, network latency and reliability, and data privacy concerns associated with API usage.
- DeepSeek-R1 (Advanced Reasoning LLM) [48]: Representing a state-of-the-art LLM known for strong reasoning capabilities, also accessed as a general-purpose model without specific fine-tuning for this task. It is generally expected to provide higher-quality reasoning but may have significantly higher inference times and shares similar practical concerns with other API-based models for real-time industrial applications.
- Ours (Full Method): The complete proposed framework, integrating the FCN for initial fault probability estimation and the LoRA fine-tuned Qwen3-4B LLM for generating an interpretable diagnostic report based on all available information.
5. Conclusions and Future Work
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Smith, W.A.; Randall, R.B. Rolling Element Bearing Diagnostics Using the Case Western Reserve University Data: A Benchmark Study. Mech. Syst. Signal Process. 2015, 64–65, 100–131. [Google Scholar] [CrossRef]
- Li, S.; Peng, Y.; Shen, Y.; Zhao, S.; Shao, H.; Bin, G.; Guo, Y.; Yang, X.; Fan, C. Rolling Bearing Fault Diagnosis Under Data Imbalance and Variable Speed Based on Adaptive Clustering Weighted Oversampling. Reliab. Eng. Syst. Saf. 2024, 244, 109938. [Google Scholar] [CrossRef]
- Dong, Y.; Li, Y.; Zheng, H.; Wang, R.; Xu, M. A New Dynamic Model and Transfer Learning Based Intelligent Fault Diagnosis Framework for Rolling Element Bearings Race Faults: Solving the Small Sample Problem. ISA Trans. 2022, 121, 327–348. [Google Scholar] [CrossRef]
- Wang, B.; Lei, Y.; Li, N.; Li, N. A Hybrid Prognostics Approach for Estimating Remaining Useful Life of Rolling Element Bearings. IEEE Trans. Reliab. 2020, 69, 401–412. [Google Scholar] [CrossRef]
- Kong, X.; Cai, B.; Yu, Y.; Yang, J.; Wang, B.; Liu, Z.; Shao, X.; Yang, C. Intelligent Diagnosis Method for Early Faults of Electric-Hydraulic Control System Based on Residual Analysis. Reliab. Eng. Syst. Saf. 2025, 261, 111142. [Google Scholar] [CrossRef]
- Shao, X.; Cai, B.; Zou, Z.; Shao, H.; Yang, C.; Liu, Y. Artificial Intelligence Enhanced Fault Prediction with Industrial Incomplete Information. Mech. Syst. Signal Process. 2025, 224, 112063. [Google Scholar] [CrossRef]
- Wang, Z.; Xuan, J.; Shi, T.; Li, Y.F. Multi-Label Domain Adversarial Reinforcement Learning for Unsupervised Compound Fault Recognition. Reliab. Eng. Syst. Saf. 2025, 254, 110638. [Google Scholar] [CrossRef]
- Wang, Z.; Xuan, J.; Shi, T. Domain Reinforcement Feature Adaptation Methodology with Correlation Alignment for Compound Fault Diagnosis of Rolling Bearing. Expert Syst. Appl. 2025, 262, 125594. [Google Scholar] [CrossRef]
- Wang, Z.; Li, S.; Xuan, J.; Shi, T. Biologically Inspired Compound Defect Detection Using a Spiking Neural Network with Continuous Time–Frequency Gradients. Adv. Eng. Inform. 2025, 65, 103132. [Google Scholar] [CrossRef]
- Wang, Z.; Xuan, J.; Shi, T. An Autonomous Recognition Framework Based on Reinforced Adversarial Open Set Algorithm for Compound Fault of Mechanical Equipment. Mech. Syst. Signal Process. 2024, 219, 111596. [Google Scholar] [CrossRef]
- Chen, X.; Zhang, B.; Gao, D. Bearing Fault Diagnosis Base on Multi-Scale CNN and LSTM Model. J. Intell. Manuf. 2021, 32, 971–987. [Google Scholar] [CrossRef]
- Smith, W.A.; Borghesani, P.; Ni, Q.; Wang, K.; Peng, Z. Optimal Demodulation-Band Selection for Envelope-Based Diagnostics: A Comparative Study of Traditional and Novel Tools. Mech. Syst. Signal Process. 2019, 134, 106303. [Google Scholar] [CrossRef]
- Chen, B.; Cheng, Y.; Cao, H.; Song, S.; Mei, G.; Gu, F.; Zhang, W.; Ball, A.D. Generalized Statistical Indicators-Guided Signal Blind Deconvolution for Fault Diagnosis of Railway Vehicle Axle-Box Bearings. IEEE Trans. Veh. Technol. 2025, 74, 2458–2469. [Google Scholar] [CrossRef]
- Antoni, J. Cyclic Spectral Analysis of Rolling-Element Bearing Signals: Facts and Fictions. J. Sound Vib. 2007, 304, 497–529. [Google Scholar] [CrossRef]
- Tao, L.; Liu, H.; Ning, G.; Cao, W.; Huang, B.; Lu, C. LLM-based Framework for Bearing Fault Diagnosis. Mech. Syst. Signal Process. 2025, 224, 112127. [Google Scholar] [CrossRef]
- Wang, Z.; Xu, Z.; Cai, C.; Wang, X.; Xu, J.; Shi, K.; Zhong, X.; Liao, Z.; Li, Q.A. Rolling Bearing Fault Diagnosis Method Using Time-Frequency Information Integration and Multi-Scale TransFusion Network. Knowl.-Based Syst. 2024, 284, 111344. [Google Scholar] [CrossRef]
- He, M.; He, D. Deep Learning Based Approach for Bearing Fault Diagnosis. IEEE Trans. Ind. Appl. 2017, 53, 3057–3065. [Google Scholar] [CrossRef]
- Bao, Z.; Du, J.; Zhang, W.; Wang, J.; Qiu, T.; Cao, Y. A Transformer Model-Based Approach to Bearing Fault Diagnosis. In Data Science: 7th International Conference of Pioneering Computer Scientists, Engineers and Educators, ICPCSEE 2021, Taiyuan, China, 17–20 September 2021, Proceedings, Part I 7; Communications in Computer and Information, Science; Zeng, J., Qin, P., Jing, W., Song, X., Lu, Z., Eds.; Springer: Singapore, 2021; pp. 65–79. [Google Scholar] [CrossRef]
- Yan, J.; Cheng, Y.; Zhang, F.; Zhou, N.; Wang, H.; Jin, B.; Wang, M.; Zhang, W. Multimodal Imitation Learning for Arc Detection in Complex Railway Environments. IEEE Trans. Instrum. Meas. 2025, 74, 1–13. [Google Scholar] [CrossRef]
- Wang, X.; Jiang, H.; Mu, M.; Dong, Y. A Dynamic Collaborative Adversarial Domain Adaptation Network for Unsupervised Rotating Machinery Fault Diagnosis. Reliab. Eng. Syst. Saf. 2025, 255, 110662. [Google Scholar] [CrossRef]
- Cheng, Y.; Yan, J.; Zhang, F.; Li, M.; Zhou, N.; Shi, C.; Jin, B.; Zhang, W. Surrogate Modeling of Pantograph-Catenary System Interactions. Mech. Syst. Signal Process. 2025, 224, 112134. [Google Scholar] [CrossRef]
- Jiang, F.; Kuang, Y.; Li, T.; Zhang, S.; Wu, Z.; Feng, K.; Li, W. Towards Enhanced Interpretability: A Mechanism-Driven Domain Adaptation Model for Bearing Fault Diagnosis across Operating Conditions. Mech. Syst. Signal Process. 2025, 225, 112244. [Google Scholar] [CrossRef]
- An, Z.; Li, S.; Wang, J.; Jiang, X. A Novel Bearing Intelligent Fault Diagnosis Framework under Time-Varying Working Conditions Using Recurrent Neural Network. ISA Trans. 2020, 100, 155–170. [Google Scholar] [CrossRef] [PubMed]
- Rai, V.; Mohanty, A. Bearing Fault Diagnosis Using FFT of Intrinsic Mode Functions in Hilbert–Huang Transform. Mech. Syst. Signal Process. 2007, 21, 2607–2615. [Google Scholar] [CrossRef]
- Silva, A.; González-Guillén, C.; Ruiz-García, M.; Domínguez-Jiménez, M.E. Improving the Sensitivity of Early Rub Detection in Rotating Machines with an Adaptive Orthogonal Filter. Mech. Syst. Signal Process. 2022, 171, 108900. [Google Scholar] [CrossRef]
- Wang, Z.; Lu, C.; Wang, Z.; Liu, H.; Fan, H. Fault Diagnosis and Health Assessment for Bearings Using the Mahalanobis–Taguchi System Based on EMD-SVD. Trans. Inst. Meas. Control 2013, 35, 798–807. [Google Scholar] [CrossRef]
- Fernández-Francos, D.; Martínez-Rego, D.; Fontenla-Romero, O.; Alonso-Betanzos, A. Automatic Bearing Fault Diagnosis Based on One-Class ν-SVM. Comput. Ind. Eng. 2013, 64, 357–365. [Google Scholar] [CrossRef]
- Xu, J.; Liang, S.; Ding, X.; Yan, R. A Zero-Shot Fault Semantics Learning Model for Compound Fault Diagnosis. Expert Syst. Appl. 2023, 221, 119642. [Google Scholar] [CrossRef]
- Shao, X.; Kim, C.S. Adaptive Multi-Scale Attention Convolution Neural Network for Cross-Domain Fault Diagnosis. Expert Syst. Appl. 2024, 236, 121216. [Google Scholar] [CrossRef]
- Wei, Q.; Tian, X.; Cui, L.; Zheng, F.; Liu, L. WSAFormer-DFFN: A Model for Rotating Machinery Fault Diagnosis Using 1D Window-Based Multi-Head Self-Attention and Deep Feature Fusion Network. Eng. Appl. Artif. Intell. 2023, 124, 106633. [Google Scholar] [CrossRef]
- Zafar, M.R.; Khan, N. Deterministic Local Interpretable Model-Agnostic Explanations for Stable Explainability. Mach. Learn. Knowl. Extr. 2021, 3, 525–541. [Google Scholar] [CrossRef]
- Wang, H.; Liu, Z.; Peng, D.; Zuo, M.J. Interpretable Convolutional Neural Network with Multilayer Wavelet for Noise-Robust Machinery Fault Diagnosis. Mech. Syst. Signal Process. 2023, 195, 110314. [Google Scholar] [CrossRef]
- Han, Y.; Lv, S.; Huang, Q.; Zhang, Y. AMCW-DFFNSA: An Interpretable Deep Feature Fusion Network for Noise-Robust Machinery Fault Diagnosis. Knowl.-Based Syst. 2024, 301, 112361. [Google Scholar] [CrossRef]
- Chen, Z.; Gryllias, K.; Li, W. Intelligent Fault Diagnosis for Rotary Machinery Using Transferable Convolutional Neural Network. IEEE Trans. Ind. Inform. 2020, 16, 339–349. [Google Scholar] [CrossRef]
- Gu, Z.; Zhu, B.; Zhu, G.; Chen, Y.; Tang, M.; Wang, J. AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models. Proc. AAAI Conf. Artif. Intell. 2024, 38, 1932–1940. [Google Scholar] [CrossRef]
- OpenAI. GPT-4 Technical Report. arXiv 2024, arXiv:2303.08774. [Google Scholar] [CrossRef]
- Li, Y.F.; Wang, H.; Sun, M. ChatGPT-like Large-Scale Foundation Models for Prognostics and Health Management: A Survey and Roadmaps. Reliab. Eng. Syst. Saf. 2024, 243, 109850. [Google Scholar] [CrossRef]
- Peng, H.; Liu, J.; Du, J.; Gao, J.; Wang, W. BearLLM: A Prior Knowledge-Enhanced Bearing Health Management Framework with Unified Vibration Signal Representation. Proc. AAAI Conf. Artif. Intell. 2025, 39, 19866–19874. [Google Scholar] [CrossRef]
- Case Western Reserve University. 2019. Available online: https://engineering.case.edu/bearingdatacenter/download-data-file (accessed on 16 June 2025).
- Lessmeier, C.; Kimotho, J.K.; Zimmer, D.; Sextro, W. Condition Monitoring of Bearing Damage in Electromechanical Drive Systems by Using Motor Current Signals of Electric Motors: A Benchmark Data Set for Data-Driven Classification. PHM Soc. Eur. Conf. 2016, 3. [Google Scholar] [CrossRef]
- Zhang, W.; Peng, G.; Li, C.; Chen, Y.; Zhang, Z. A New Deep Learning Model for Fault Diagnosis with Good Anti-Noise and Domain Adaptation Ability on Raw Vibration Signals. Sensors 2017, 17, 425. [Google Scholar] [CrossRef]
- He, K.; Zhang, X.; Ren, S.; Sun, J. Deep Residual Learning for Image Recognition. In Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 27–30 June 2016; pp. 770–778. [Google Scholar] [CrossRef]
- Qwen Team. Qwen3 Technical Report. arXiv 2025, arXiv:2505.09388. [Google Scholar]
- Hu, E.J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; Chen, W. LoRA: Low-Rank Adaptation of Large Language Models. arXiv 2021, arXiv:2106.09685. [Google Scholar] [CrossRef]
- Mangrulkar, S.; Gugger, S.; Debut, L.; Belkada, Y.; Paul, S.; Bossan, B. PEFT: State-of-the-art Parameter-Efficient Fine-Tuning Methods. 2022. Available online: https://github.com/huggingface/peft (accessed on 16 June 2025).
- Liao, J.X.; Dong, H.C.; Sun, Z.Q.; Sun, J.; Zhang, S.; Fan, F.L. Attention-Embedded Quadratic Network (Qttention) for Effective and Interpretable Bearing Fault Diagnosis. IEEE Trans. Instrum. Meas. 2023, 72, 1–13. [Google Scholar] [CrossRef]
- He, Z.; Shao, H.; Zhong, X.; Zhao, X. Ensemble Transfer CNNs Driven by Multi-Channel Signals for Fault Diagnosis of Rotating Machinery Cross Working Conditions. Knowl.-Based Syst. 2020, 207, 106396. [Google Scholar] [CrossRef]
- DeepSeek-AI. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. arXiv 2025, arXiv:cs.CL/2501.12948.
Layer Name | Input Channels | Output Channels | Kernel Size/Stride/Padding | Activation | Other Operations |
---|---|---|---|---|---|
Input | 1 | - | - | - | Normalized (1 × 12,000) |
Wide Layer | 1 | 36 | 32/4/15 | LeakyReLU | BatchNorm1D |
MSM Block 1 | 36 | 108 | Multi-Scale (3,5,7)/2, Shortcut (1)/1 | ReLU | Attention, BatchNorm, Residual |
MSM Block 2 | 108 | 216 | Multi-Scale (3,5,7)/2, Shortcut (1)/1 | ReLU | Attention, BatchNorm, Residual |
MSM Block 3 | 216 | 216 | Multi-Scale (3,5,7)/2, Shortcut (1)/1 | ReLU | Attention, BatchNorm, Residual |
Global Pooling | 216 | 432 | - | - | AdaAvgPool, AdaMaxPool, Concatenate |
FC Layer 1 | 432 | 128 | - | ReLU | Linear |
FC Layer 2 (Output) | 128 | 10 | - | Softmax | Linear |
Model | Accuracy | Precision | F1-Score | FPR | FNR | MSE |
---|---|---|---|---|---|---|
CNN | 0.8653 | 0.6359 | 0.6325 | 0.0139 | 0.1347 | 0.0178 |
WDCNN | 0.8565 | 0.5971 | 0.5866 | 0.0161 | 0.1738 | 0.0178 |
MCNN | 0.8634 | 0.6540 | 0.6431 | 0.0165 | 0.1387 | 0.0216 |
TCNN | 0.9317 | 0.7117 | 0.7051 | 0.0080 | 0.0756 | 0.0097 |
QCNN | 0.9491 | 0.7000 | 0.6974 | 0.0058 | 0.0949 | 0.0068 |
SCNN | 0.9838 | 0.7651 | 0.7637 | 0.0019 | 0.0175 | 0.0028 |
FCN | 0.9919 | 0.7652 | 0.7643 | 0.0009 | 0.0053 | 0.0010 |
Method | Time (s) ± Std | Accuracy | Understand | Diagnosis Logic | Evidence Relevance | Utility | Trustworth | Overall Rating |
---|---|---|---|---|---|---|---|---|
FCN + Mapping | 0.1 ± 0.0 | 4.9 | 1.2 | 1.0 | 1.1 | 1.3 | 1.4 | 1.6 |
w/o FCN | 2.9 ± 0.5 | 3.2 | 3.8 | 3.1 | 3.3 | 3.5 | 3.4 | 3.1 |
w/o LoRA | 2.9 ± 0.5 | 4.8 | 3.5 | 3.9 | 3.7 | 4.0 | 4.1 | 4.0 |
ChatGPT 4.1 | 8.2 ± 4.3 | 4.9 | 4.8 | 4.7 | 4.5 | 4.6 | 4.7 | 4.5 |
Deepseek R1 | 28.2 ± 11.2 | 4.9 | 4.9 | 5.0 | 4.9 | 4.8 | 4.9 | 4.1 |
Ours | 3.0 ± 0.5 | 4.9 | 4.9 | 4.9 | 4.9 | 5.0 | 5.0 | 4.9 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Bao, P.; Yi, W.; Zhu, Y.; Shen, Y.; Peng, H. A Spectral Interpretable Bearing Fault Diagnosis Framework Powered by Large Language Models. Sensors 2025, 25, 3822. https://doi.org/10.3390/s25123822
Bao P, Yi W, Zhu Y, Shen Y, Peng H. A Spectral Interpretable Bearing Fault Diagnosis Framework Powered by Large Language Models. Sensors. 2025; 25(12):3822. https://doi.org/10.3390/s25123822
Chicago/Turabian StyleBao, Panfeng, Wenjun Yi, Yue Zhu, Yufeng Shen, and Haotian Peng. 2025. "A Spectral Interpretable Bearing Fault Diagnosis Framework Powered by Large Language Models" Sensors 25, no. 12: 3822. https://doi.org/10.3390/s25123822
APA StyleBao, P., Yi, W., Zhu, Y., Shen, Y., & Peng, H. (2025). A Spectral Interpretable Bearing Fault Diagnosis Framework Powered by Large Language Models. Sensors, 25(12), 3822. https://doi.org/10.3390/s25123822